位置数据中的城市行业空间特征挖掘

Spatial feature discovery of urban industry based on location data

针对城市空间特征分析问题,该文以武汉市中心城区为研究对象,对多维位置数据进行数据挖掘,获得城市行业空间分布特征模式。首先通过对聚类切分后的区域从不同维度进行切片,得到各个区域不同行业类别在多个属性切面的量化特征。再通过构建动态加权密度聚类模型,提出以非位置属性计算权重系数修正判定位置点分类的相似度函数,实现对城市重点区块的抽取。综合切片和聚类提取结果,归纳出了目标城市行业空间特征模式。仿真结果表明,基于多维位置数据的数据挖掘可以实现对城市行业空间特征的直接提取,抽取得到的重点区块具有明显的高热特性。聚类分析结果与城市行业基本特征的吻合度较高,表明所提出的动态加权聚类算法较传统的密度聚类算法更适合于多维位置数据的挖掘,同时也为现有的城市空间布局模式研究中对位置点集密度聚类分析难于设置网格大小和密度带宽的问题提供了一条新的思路。

Aimed at the problem of urban spatial feature analysis,this paper taked Wuhan city as the research object,conducted data mining and analysis on multidimensional location data to obtain the spatial distribution feature pattern of urban industry. The quantitative characteristics of each attribute section were summarized by slicing statistics of the four clustering regions from different dimensions. The dynamic weighted density clustering model was constructed,and the non-location attribute was used to calculate the weight coefficient to modify the similarity function,which realized the extraction of the key blocks of the target city,and verified the high heat characteristics of the extracted key blocks. Based on the results,the spatial characteristic pattern of target urban industries was summarized. Simulation results showed that the data mining based on multidimensional location data could realize urban industry space characteristics discovery. Comparing with the basic characteristics of city industries,the clustering analysis result of urban industry had a good fit. It showed that the proposed dynamic weighted clustering algorithm most suitable for mining multidimensional location data. It also provided a new idea for the problem that it was difficult to set grids’ size and bandwidth for density clustering in the study of urban spatial layout.